Tech Giants Fund AI Rivals, Creating Circular Capitalism

Per The Economic Times, Alphabet has committed to invest up to $40 billion in Anthropic, the company behind Claude. Anthropic's Feb 12, 2026 announcement states it raised $30 billion in a Series G at a $380 billion post-money valuation, led by GIC and Coatue, and lists many institutional investors while noting a portion of the round includes Microsoft and NVIDIA. The Economic Times reports Amazon has already invested heavily in Anthropic and has planned up to $25 billion more. The Economic Times also recounts prior reporting that Anthropic agreed to purchase $30 billion of Azure compute powered by NVIDIA hardware.

Industry reporting highlights a loop where capital injected into frontier AI companies often returns to large cloud providers and chip makers as infrastructure spend. For practitioners, this reinforces that large-model development remains tightly coupled to rented GPU/TPU capacity and specialised accelerators. Companies training at scale typically convert investment dollars into months or years of cloud and on-premise compute consumption, creating concentrated demand spikes for specific instance types and interconnects.

Observed patterns in similar funding and procurement cycles show three effects:

• •reported cloud revenue can grow when investors also buy compute from those same cloud vendors
• •chip vendors extract value both via direct equity stakes and hardware sales
• •enterprise customers face shifting procurement leverage as hyperscalers secure long-term capacity commitments. These are generic industry patterns, not claims about Anthropic or its investors beyond the sourced transactions

Indicators to follow include renewal and minimum-spend terms in cloud purchase agreements, public disclosures about committed-capacity vs spot usage, how investors disclose related-party compute arrangements in filings, and any regulatory or accounting scrutiny of unusually circular capital flows. Practitioners should monitor price and availability trends for the GPU/TPU instance families that power large-model training and inference.